GPU component level diagnosis and repair service?

[jdempsey]

Hey guys, I've got a 1080ti that's on the fritz, and while I haven't finished testing every scenario I can think of, I'm beginning to suspect its an issue with some smd, cap, or potentially a vram chip.

It works intermittently, but every time I've removed the card, put it back in the system it's from, or others, I get different weirdness. Visual inspection doesn't show any obvious signs of damaged components, etc. but symptoms will range from whatever system it's in either not booting, rebooting repeatedly, video artifacting, driver crashes, etc, but they're usually consistent (repeatable), with whichever symptom it's showing.

Yet at one point, after problems first started manifesting, and i thought it was toast, I got it to work fine for a week (even in gaming and under stress without a single hiccup), before I moved the system off the bench back to where it usually lives, and it was back to it.

I've checked for shorts, even removed every component from that system and checked each in isolation, and the gpu in other machines with different PSU's and on different MBs.


I mean, I will probably go ahead and buy another card regardless, but I kind of hate to give up on it, if it's a flakey cap, or something similar though should be easy enough to replace, although this is a little outside of my wheel house (I do have a background in this stuff, just not at the circuit level). This card was a pain from day one, being an MSI Armor OC (the one that had all the overheating issues that MSI was refusing to acknowledge), that I ended up watercooling, to great results in terms of the OC performance, since at the time, getting a replacement was impossible with the mining scene. I'm kind of stubbornly invested in not letting go until there's no hope with this thing.



Figured I'd ask here though, if there's any outfit, or individual, that takes on this kind of work, before I start poking around a bit more invasively with the DMM and scope. Anyway, thanks for the help guys!

 

[hititnquitit]

Have you pulled the block off yet? If so anything look unusual like white marks on the pcb etc? Or any black marks on the block?

 

[jdempsey]

Have you pulled the block off yet? If so anything look unusual like white marks on the pcb etc? Or any black marks on the block?

Yeah, that's my last step before I start testing every SMD for a short to ground. I pulled it off, and just gave it a cursory, but haven't cleaned it up yet.

What's the significance of white marks? It's a black PCB I've seen some stuff on the back side that almost looks like powdery galvanic corrosion, even though nothing was mounted there, but it could just be shit from my environment. Lots of machinery in here and it's and old dusty building. Also, in fairness to the card, I did cart that whole system up to my remote mountain cabin a few times, up my road that's 2 track, 4 wheel drive only, that people would likely pay to wheel on. So there was plenty of bouncing around and I should have honestly removed the card before taking it, considering how heavy it is.

Originally it was using a riser (I've got it in one of those open Thermaltake cases), so I thought it was toast, since I pulled it and got it working again. Then I thought maybe the PCI-E slot was crapping out, but it seems ok. I've tested all the voltages on the PSU also, but like I said, similar issues with the card in another machine.



I highly suspect an electrical component on the card itself. So I'll double check for swollen caps or any component that's grounding on both legs.

 

[hititnquitit]

The white may show the residue of a small cooked component.
Sounds like it was on borrowed time with all of the travel it endured!

 

[jdempsey]

The white may show the residue of a small cooked component.
Sounds like it was on borrowed time with all of the travel it endured!

Yeah very possible, but I'm stubborn, and with the prices these devices command these days, it seems kind of crazy how consumable we treat them. No reason they shouldn't be serviceable. Although that's mostly finding someone qualified and willing to do the work I guess. That said, I know I'd be willing to pay an extra cost to have a GPU designed for easier component replacement. Who wouldn't want say, socketed vram for example, push it to the limit OC'ing, and you could pull and replace at will. Ironically, if you look at (or remember) the way computer hardware was built in the early days, almost every component was easily replaceable. In fairness, it probably wasn't that heavily utilized, and driving costs lower necessitated leaner manufacturing, but considering the ever increasing prices of top tier video cards especially, you'd think someone would get the idea again, especially for the higher end offerings geared toward heavy OCing. Seems like a no brainer? Lol, sorry, rambling again.